------- Comment #5 from martin dot drab at fjfi dot cvut dot cz 2008-04-27 22:54 ------- And the test cases also fail with
-O1 -finline-small-functions -ftree-vrp -m32 so if you compile it for 32-bit target, naither the -funsafe-loop-optimizations nor the -fprefetch-loop-arrays needs to be specified. Haven't tested it on a native 32-bit host, so, hard to say whether it would ICE there as well, just like this. However, natively on a 64-bit host with just "-O1 -finline-small-functions -ftree-vrp" it works without an ICE. -- http://gcc.gnu.org/bugzilla/show_bug.cgi?id=36066
